A&A517,A94(2010) Astronomy DOI:10.1051/0004-6361/201014566 & (cid:2)c ESO2010 Astrophysics (cid:2) The clusters Abell 222 and Abell 223: a multi-wavelength view F.Durret1,2,T.F.Laganá3,C.Adami4,andE.Bertin1,2 1 UPMCUniversitéParis06,UMR7095,Institutd’AstrophysiquedeParis,75014Paris,France e-mail:[email protected] 2 CNRS,UMR7095,Institutd’AstrophysiquedeParis,75014Paris,France 3 IAG,USP,R.doMatão1226,05508-090.SãoPaulo/SP,Brazil 4 LAM,Pôledel’ÉtoileSitedeChâteau-Gombert,38rueFrédéricJoliot-Curie,13388MarseilleCedex13,France Accepted30March2010/Received7May2010 ABSTRACT Context.TheAbell222 and223 clustersarelocatedat anaverageredshift z ∼ 0.21 andareseparatedby 0.26deg. Signaturesof mergershavebeenpreviouslyfoundintheseclusters,bothinX-raysandatopticalwavelengths,thusmotivatingourstudy.InX-rays, they arerelativelybright, and Abell223 shows adouble structure. A filament hasalso beendetected between theclustersboth at opticalandX-raywavelengths. Aims.Weanalyse theopticalproperties ofthesetwoclustersbasedon deepimaging intwobands, derivetheir galaxyluminosity functions(GLFs)andcorrelatethesepropertieswithX-raycharacteristicsderivedfromXMM-Newtondata. Methods.TheopticalpartofourstudyisbasedonarchiveimagesobtainedwiththeCFHTMegaprime/Megacamcamera,coveringa totalregionofabout1deg2,or12.3×12.3Mpc2ataredshiftof0.21.TheX-rayanalysisisbasedonarchiveXMM-Newtonimages. Results.The GLFs of Abell 222 in the g(cid:4) and r(cid:4) bands are well fit by a Schechter function; the GLF is steeper in r(cid:4) than in g(cid:4). ForAbell 223, theGLFsinbothbands requireasecond component atbright magnitudes, added toaSchechter function; theyare similarinbothbands. TheSerna&Gerbalmethod allowstoseparatewellthetwoclusters.Noobvious filamentarystructuresare detectedatverylargescalesaroundtheclusters,butathirdclusteratthesameredshift,Abell209,islocatedataprojecteddistanceof 19.2Mpc.X-raytemperatureandmetallicitymapsrevealthatthetemperatureandmetallicityoftheX-raygasarequitehomogeneous inAbell222,whiletheyareveryperturbedinAbell223. Conclusions.TheAbell222/Abell223systemiscomplex.Thetwoclustersthatformthisstructurepresentverydifferentdynamical states. Abell 222 is a smaller, less massive and almost isothermal cluster. On the other hand, Abell 223 is more massive and has most probably been crossed by a subcluster on its way to the northeast. As a consequence, the temperature distribution is very inhomogeneous. Signs of recent interactions are also detected in the optical data where this cluster shows a “perturbed” GLF. In summary,themultiwavelengthanalysesofAbell222andAbell223areusedtoinvestigatetheconnectionbetweentheICMandthe clustergalaxypropertiesinaninteractingsystem. Keywords.galaxies:clusters:individual:Abell222–X-rays:galaxies:clusters–galaxies:clusters:individual:Abell223– galaxies:clusters:intraclustermedium 1. Introduction Bouéetal.(2008),thisslopecanstronglyvaryfromonecluster toanother,andfromoneregiontoanotherinagivencluster. Galaxy evolution is known to be influenced by environmen- Supersedingthe dynamicalanalyses of Proust et al. (2000) tal effects, which are particularly strong in galaxy clusters, all and Girardi & Mezzetti (2001), Dietrich et al. (2002) have ob- the more when they are undergoingmerging events. The anal- tained a redshift catalogue for 183 galaxies in the region of ysis of galaxyluminosityfunctions(GLF), and in particular of the cluster pair Abell222/223,outof which153belongto one their faint-end slopes is a good way to trace the history of the of the two clusters. Their analysis was also based on photom- faint galaxypopulationand the influence of mergingeventson etry in the V and R bands. They have estimated redshifts of this population. As summarized for example in Table A.1 of 0.2126±0.0008and0.2079±0.0008,andvelocitydispersionsof 1014and1032kms−1,forAbell222andAbell223respectively. (cid:2) BasedonobservationsobtainedwithMegaPrime/MegaCam,ajoint AsindicatedbyDietrichetal.(2002)thesevelocitydispersions project of CFHT and CEA/DAPNIA, at the Canada-France-Hawaii are somewhathigherthanthose expectedfromthe X-raylumi- Telescope(CFHT)whichisoperatedbytheNationalResearchCouncil nosities.Abell222appearsrelativelyrelaxed,withagalaxyve- (NRC)ofCanada,theInstitutNationaldesSciencesdel’Universofthe locityhistogramconsistentwithaGaussian.Ontheotherhand, Centre National de la Recherche Scientifique (CNRS) of France, and theadaptivekernelgalaxydensitymapofAbell223showstwo the University of Hawaii. This work has also made use of data prod- peaks,thoughsurprisinglytheDressler-Schectmantest(Dressler uctsproducedatTERAPIXandtheCanadianAstronomyDataCentre & Schectman1988) doesnotfindthese two peaksandtheDIP aspartoftheCanada-France-Hawaii TelescopeLegacySurvey,acol- statistics cannotrejectthe nullhypothesisofa unimodaldistri- laborative project of NRC and CNRS. The X-ray analysis is based bution, implying that a Gaussian parent population cannot be on XMM-Newton archive data. This research has made use of the NASA/IPACExtragalacticDatabase(NED)whichisoperatedbytheJet excluded.Dietrichetal.(2002)derivedrespectivemasstolight Propulsion Laboratory, CaliforniaInstituteof Technology, under con- ratiosof202±43and149±33h70 M(cid:5)/L(cid:5)andluminosityfunc- tractwiththeNationalAeronauticsandSpaceAdministration. tionsintheR banddowntoabsolutemagnitudesof−19.5,and ArticlepublishedbyEDPSciences Page1of15 A&A517,A94(2010) Table1.Summaryoftheobservations. Filter g(cid:4) r(cid:4) Numberofcoaddedimages 5 8 Exposuretime(s) 2000 4000 Seeing(arcsec) 0.81 0.73 Limitingmagnitude(5σ) 26.7 26.4 found thata bridgeof galaxiesseems to exist between the two clusters.Notethatsuchbridgesofgalaxiesjoiningtwoclusters arestillquiterare,anotherexampleatacomparableredshiftbe- ingthefilamentdetectedbetweenAbell1763andAbell1770by Faddaetal.(2008). A weak lensing analysis of the Abell 222/223 system was performed by Dietrich et al. (2005), who found evidence for a possible dark matter filament connecting both clusters. They interpret the difference in redshift between Abell 222 and Abell 223 as implying a separation along the line of sight of 15±3h−1 Mpc. Their weak lensing surface mass density con- 70 toursagainshowevidenceforadoublestructureinAbell223. Theexistenceofa1.2Mpcwidefilamentbetweenbothclus- terswasalsofoundinX-raysbyWerneretal.(2008). Fig.1.Centralsurfacebrightnessinther(cid:4)bandasafunctionofr(cid:4)mag- nitude. Thehorizontal andoblique fullredlinesseparate thegalaxies We presenthere a muti-wavelengthopticalandX-ray anal- (abovetheselines)fromthestars(belowtheselines).Theobliquedot- ysisoftheclustersAbell222andAbell223.Theopticalpartis ted red line shows the limit for stars computed for 17.25 < r(cid:4) < 22 basedonimagingintheg(cid:4)andr(cid:4)bandsobtainedwiththeCFHT (see text). The two vertical linescorrespond to r(cid:4) = 17.25 where the Megaprime/Megacamcameraandretrievedfromthe Megapipe starsstopbeingsaturated,andr(cid:4)=22,wherethefittocalculatethestar image stacking pipeline at CADC (Gwyn 2009). As seen in sequencewaslimited(seetext). Table 1, the r(cid:4) band data are about 0.55 mag deeper than the g(cid:4)banddata(bytakinganaveragecolorg(cid:4)−r(cid:4) =0.85),andwill begivenpriorityinthemorphologicalanalysis;theg(cid:4) datawill We did not use the catalogues available for these images, be used to select brightcluster membersfromcolor-magnitude because they were made without masking the surroundings of diagrams,andalsobecausetheg(cid:4) bandismoresensitivetostar bright stars, so we preferred to build masks first, then to ex- formationthanther(cid:4) band.Theseimagesaredeeperthanthose tract sourceswith SExtractor (Bertin & Arnouts1996). Due to ofDietrichbyatleast0.5mag(ourrawgalaxycountsstartde- theditheringpattern,thetotalareacoveredbyther(cid:4) imagewas creasingforr(cid:4) > 24.5whileDietrichetal.saythattheircounts slightlylargerthantheMegacam1deg2field:1.196deg2. intheRbandstopfollowingapowerlawforR>24),andthere- Objectsweredetectedandmeasuredinthefullr(cid:4)image,then fore allow to probe the GLF to somewhat fainter magnitudes. measured in the g(cid:4) image in double image mode. Magnitudes OurX-rayanalysisisbasedonarchiveXMM-Newtondata. areintheABsystem.Theobjectslocatedinthemaskedregions Wewillconsiderhereafterthatthetwoclustershaveacom- werethentakenoutofthecatalogue,leadingtoafinalcatalogue mon redshiftof 0.21,for whichWright’s cosmologycalculator of 223414 objects. All these objects have measured r(cid:4) magni- (Wright 2006) givesa distance of 1035Mpc, a spatial scale of tudes,and214669alsohavemeasuredg(cid:4)magnitudes. 3.427 kpc/arcsecand a distance modulusof 40.07(assuming a Since the r(cid:4) catalogueis deeper,and the seeing is also bet- flatΛCDMcosmologywith H = 70kms−1 Mpc−1,Ω = 0.3 ter in this band (see Table 1), we will perform our star-galaxy 0 M andΩΛ =0.7). separationinther(cid:4)band. The paper is organised as follows. We describe our optical analysis in Sect. 2, and results concerningthe galaxy luminos- 2.2.Star-galaxyseparation ity functionin Sect. 3. A search for substructuresin bothclus- ters based on the Serna-Gerbalmethod is described in Sect. 4. Inordertoseparatestarsfromgalaxies,weplottedthemaximum The X-ray data analysis andresults, includingtemperatureand surface brightness μ in the r(cid:4) band as a function of r(cid:4). The max metallicitymaps,arepresentedinSect.5.Anoverallpictureof resultisshowninFig.1. thisclusterpairisdrawninSect.6. The bestfit to the star sequencevisible in Fig. 1 calculated for 17.25 < r(cid:4) < 22 is μ = 0.996∗r(cid:4) −0.247, with stan- max darddeviationsontheslopeandconstantof0.002and0.047re- 2. Opticaldataandanalysis spectively. The point-source (hereafter called “star”) sequence is clearly visible for r(cid:4) < 22, with the star saturation show- 2.1.Theopticaldata ing well for r(cid:4) < 17.25. We will define galaxies as the ob- We have retrieved from the CADC Megapipe archive (Gwyn jects with μ (r(cid:4)) > 17.25 for r(cid:4) < 17.25, and as the objects max 2009) the reduced and stacked images in the g(cid:4) and above the line of equation μ = 0.996 ∗ r(cid:4) − 0.247 + 0.3, max r(cid:4) bands (namely G002.024.493-12.783.G and G002.024.493- or μ = 0.996∗ r(cid:4) + 0.053. Stars will be defined as all the max 12.783.R) and give a few details on the observations in otherobjects(seeFig.1).Thesmallcloudofpointsobservedin Table 1. Observations were made at the CFHT with the Fig. 1 under the star sequence is in fact defects, but represents Megaprime/Megacamcamera,whichhasapixelsizeof0.186× lessthan2%ofthenumberofstars.Wethusobtainastaranda 0.186arcsec2. galaxycatalogue. Page2of15 F.Durretetal.:Abell222/223 As a checkto see up to whatmagnitudewe couldtrustour star-galaxyseparation,we retrievedthe star cataloguefromthe BesançonmodelforourGalaxy(Robinetal.2003)ina1deg2 regioncenteredonthepositionoftheimageanalysedhere.Such a catalogue is in AB magnitudes(as ours) and is corrected for extinction. In order for it to be directly comparable to our star catalogue,wecorrectedourstarcatalogue(andourgalaxycata- logueaswell,forlaterpurposes)forextinction:0.071maging(cid:4) and 0.052 mag in r(cid:4) (as derived from the Schlegel et al. 1998 maps). Ther(cid:4) magnitudehistogramoftheobjectsclassifiedasstars inourr(cid:4)imageroughlyagreeswiththeBesançonstarcatalogue for r(cid:4) ≤ 20, though in some bins our counts tend to be some- what higher than those of the Besançon model even for bright magnitudes.Forr(cid:4) >20,wedetectmorestarsthanpredictedby the Besançon model (the difference is about 35% at r(cid:4) ∼ 21). Thiscouldimplythatourstar-galaxyseparationisnotvalidfor r(cid:4) > 20;however,assuggestedbyFig. 1,thisisaratherbright limit, as confirmed by simulations adapted to match compara- ble data, where we found that the star-galaxy separation deter- minedwiththismethodwasreliableatleastuptor(cid:4) =21(Boué Fig.2.Positionsoftheobjects(starsandgalaxies)intheAbell222(ma- genta) and Abell 223 (cyan) catalogues. The objects used to estimate etal.2008).Inreality,thedisagreementbetweenourstarcounts backgroundcounts(seetext)areshowninblueforthenorthrectangle and those predicted by the Besançon model is most probably andinredfortheannulussurroundingtheclusters.Notethatthefigure duetothefactthatourfieldislocatedinthedirectionofoneof covers1.2×1.2deg2,andisslightlylargerthantheimages,whichcover the densest regionsof the Sagittarius stream (Ibata et al. 2001; 1.04×1.15deg2. Yannyetal. 2009).Itisthereforenormalto observemorestars thanpredictedbytheBesançonmodel,whichdoesnottakethis streamintoaccount.Notethattheremayalsobeasmallcontri- butionduetoquasarsandAGN,sincetheirnumberatr(cid:4) ∼ 20.5 the CCD images and then attempt to recover them by running SExtractor again with the same parameters used for object de- isexpectedtobeabout20persquaredegree(seeRichardsetal. tectionandclassificationontheoriginalimages.Inthisway,the 2006,Fig.13). For r(cid:4) > 20, we will conservatively consider that the star- completenessismeasuredontheoriginalimages. Inpractice,weextractfromthefullfieldofviewtwosubim- galaxy separation may be wrong, and we will compute galaxy ages, each 1500×1500pixels2, correspondingto the positions counts by counting the total number of objects (galaxies plus oftheAbell222andAbell223clustersontheimage. stars) per bin of 0.5 mag, and considering that the number of Ineachsubfield,andforeach0.5magbinbetweenr(cid:4) = 20 galaxiesisequaltothetotalnumberofobjectsminusthenumber and27,wegenerateandaddtotheimageonestarthatwethen ofstarspredictedineachbinbytheBesançonmodel.Asmen- try to detectwith SExtractor,assumingthe same parametersas tionedbelow,wehavecheckedthatthegalaxynumbercountsin the20<r(cid:4) <22rangeestimatedbybothmethods(i.e.ourstar- previously. This process is repeated 100 times for each of the twofieldsandbands. galaxyseparationandthesubtractionoftheBesançoncountsto thetotalnumberofobjects)areconsistentwithinerrorbars. Such simulations give a completeness percentage for stars. Since the clusters are quite distant, they do not cover the This is obviously an upper limit for the completenesslevel for wholefield,soweextractedfromthestarandgalaxycatalogues galaxies,sincestarsareeasiertodetectthangalaxies.However, two catalogues as large as possible corresponding to the two wehaveshowninapreviouspaperthatthismethodgivesagood clusters. We defined the J2000.0 positions of Abell 222 and estimate of the completenessfornormalgalaxiesif we applya Abell223ascoincidingwiththe brightestgalaxyofeachclus- shift of ∼0.5 mag (see Adami et al. 2006). Results are shown ter,i.e.24.3921,−12.9912and24.5099,−12.7575respectively inFig.3. (indegrees).Notethatthesearenotexactlytheclusterpositions Fromthesesimulations,andtakingintoaccountthefactthat givenbytheNEDdatabase.Themaximumpossibleradiustoob- resultsareworseby∼0.5magformeangalaxypopulationsthan tainindependentcataloguesforthetwoclusterswas0.1290deg, forstars,wecanconsiderthatourgalaxycatalogueiscomplete or1.6Mpcataredshiftof0.21. tobetterthan80%forg(cid:4) ≤25andr(cid:4) ≤24.5. FortheAbell222andAbell223clustersseparately,wethus obtainedthreecomplementarycatalogues(withg(cid:4) andr(cid:4) mag): 2.4.Galaxycounts objects classified as galaxies (classification valid at least for r(cid:4) ≤ 20),objectsclassifiedasstars,andacompletecatalogueof The surfacescoveredby the Abell222and223catalogues(af- galaxies+starswhichwillbeusedforr(cid:4) > 20.Thepositionsof terexcludingmaskedregions)are0.0429deg2and0.0475deg2 thegalaxiesintheregionsofthetwoclustersareshowninFig.2. respectively. Galaxy counts were computed in bins of 0.5 mag normalizedtoasurfaceof1deg2. For r(cid:4) ≤ 20, galaxy counts were derived directly by com- 2.3.Cataloguecompleteness puting histograms of the numbers of galaxies in the Abell 222 Thecompletenessofthecatalogueisestimatedbysimulations. andAbell223catalogues.Forr(cid:4) > 20,webuiltforeachcluster For this, we add “artificial stars” (i.e. 2-dimensional Gaussian histogramsofthetotalnumbersofobjects(galaxies+stars)and profiles with the same Full-Width at Half Maximumas the av- obtainedgalaxycountsbysubtractingthenumbersofstarspre- erage image Point Spread Function) of different magnitudesto dictedbytheBesançonmodel.Theresultinggalaxycountswill Page3of15 A&A517,A94(2010) Fig.3.Pointsourcecompletenessasafunctionofmagnitudeinpercent- agesforAbell222(top)andAbell223(bottom)ing(cid:4)(left)andr(cid:4)(right) forpoint-likeobjects(seetext). beusedinthenextsectiontoderivetheGLFsforbothclusters inbothbands. As a test, we considered the galaxy counts in the four 0.5 mag bins for 20 < r(cid:4) < 22 computed for both clusters with the two methods (i.e. first method: considering that the star-galaxy separation is valid, and second method: consider- ingthetotalnumberofobjects(galaxies+stars)andsubtracting the number of stars predicted by the Besançon model to ob- tain the number of galaxies). In all cases, the differences are smaller than 15% (and in most cases they are only a few per- cent). Therefore, the limit (between r(cid:4) = 20 and 22) at which weconsiderthatourstar-galaxyseparationisnotvalidanymore willnotstronglyinfluenceourresults. Note that no k-correction was applied to the galaxy magnitudes. Fig.4.(g(cid:4)−r(cid:4))vs.r(cid:4) colour−magnitudediagramsforAbell222(top) 3. Results:colour–magnitudediagramsandgalaxy and Abell 223 (bottom) for objects classified as galaxies from the luminosityfunctions μmax-magnitude relation. The vertical dashed lines indicate the mag- nitude interval where the colour−magnitude relation was computed. Inorderto computethe galaxyluminosityfunctionsofthetwo The long oblique line shows the colour−magnitude relation, while clusters,itisnecessarytosubtracttothetotalgalaxycountsthe the two short oblique lines indicate an interval of ±3σ around the numbercountscorrespondingtothecontaminationbythefore- colour−magnituderelation.Thefilledcirclesshowthegalaxiesbelong- groundandbackgroundgalaxies. ingtotheclustersaccordingtotheirspectroscopicredshifts. For galaxies brighter than r(cid:4) = 20 we will select galaxies with a high probability to belong to the clusters by drawing colour-magnitudediagramsandselectinggalaxieslocatedclose beabsolutelyidentical).Wetheneliminatedthegalaxieslocated tothisrelation.Afewspiralsmaybemissedinthisway,buttheir morethan3σawayfromthisrelationandcomputedtheg(cid:4)−r(cid:4) numberinanycase isexpectedtobe small,asexplainedatthe vs.r(cid:4)relationsagain. end of Sect. 3.1 (also see e.g. Adamiet al. 1998). For galaxies The equations of the colour–magnituderelations are found fainterthanr(cid:4) =20wewillsubtractgalaxycountsstatistically. tobe:g(cid:4)−r(cid:4) =−0.0545r(cid:4)+2.091with3σ=0.21forAbell222, andg(cid:4)−r(cid:4) =−0.0474r(cid:4)+1.945with3σ=0.21forAbell223. Forr(cid:4) < 20,we willconsiderhereafterthatallthegalaxies 3.1.Colour–magnitudediagrams locatedwithin3σoftheserelations(i.e.betweenthetwoblack Themeanvaluesforg(cid:4)−r(cid:4)are0.85and0.87forAbell222and linesofFig.4)belongtotheclusters.Withthiscondition,there Abell 223. The g(cid:4) − r(cid:4) vs. r(cid:4) colour−magnitude diagrams are are141and144galaxiesbelongingtoAbell222andAbell223 showninFig.4forthetwoclusters.Asequenceiswelldefined respectivelywithr(cid:4) ≤20. forgalaxiesinthemagnituderange18<r(cid:4) <22inbothclusters. We also plot in Fig. 4 the galaxies belonging to the clus- We computed the best fit to the g(cid:4) − r(cid:4) vs. r(cid:4) relations in this ters according to their spectroscopic redshifts, i.e. galaxies in magnituderangebyapplyingalinearregression(separatelyfor the[0.195,0.215]redshiftinterval(seeSect.4).Wecanseethat thetwoclusters,sincetheircolour-magnituderelationsmaynot theirpositionsagreewellwiththecolour−magnitudeselection. Page4of15 F.Durretetal.:Abell222/223 Forbothclusters,wehaveestimatedthenumberofblueclus- tergalaxieslostbyselectinggalaxiesintheredsequenceinterval inthefollowingway.First,wecomputedhistogramsofnumbers of galaxies in the red sequence and below the red sequence in bins of 1 absolute magnitudein the r(cid:4) band. The bins of inter- esthereare Mr(cid:4) =−21.5and−20.5(seeFig.4),roughlycorre- spondingtor(cid:4) = 18.5and19.5.Thenweestimatedthenumber offoregroundgalaxiesexpected.Sincethecomovingvolumeat z = 0.21is2.622Gpc3 (Wright2006),andeachofourclusters coversan areaof0.0522deg2 onthesky,thevolumeinthe di- rectionofeachclusteris3318Mpc3.ByusingtheRbandlumi- nosityfunctionbyIlbertetal.(2005)inthe[0.05−0.20]redshift bin (see their Fig. 6 and Table 1), we find that the percentages of“lost”galaxiesareoftheorderof10%−25%forMr(cid:4) =−20.5 andof10%forMr(cid:4) =−21.5. 3.2.Comparisonfield In order to perform a statistical subtraction of the background contributionforr(cid:4) >20,weconsideredseveralpossibilities. The most obvious solution would be to extract the back- Fig.5.Galaxycountsintheg(cid:4)(left)andr(cid:4)(right)bandsformagnitudes groundinanannulussurroundingtheclusters.Thelargestouter r(cid:4) > 20wherethebackground mustbesubtractedstatistically,inlog- arithmicscale.ThecountsinAbell222and223aredrawninmagenta circlethatcanbeextracted,centeredonthemiddlepositionbe- tweenthetwoclusters(24.4510,−12.8744J2000.0indegrees), andcyanrespectively. Theblackdashedlinesshowthegalaxycounts fromtheCFHTLSDeepsurvey,andthefullgreenlineshowstheaver- hasaradiusof0.4472deg,or5.55Mpc.Wecanthentakeanin- ageoftheDeep1,Deep2andDeep3fieldcounts. Theblueandred nercircleofradius0.4028deg,or5Mpc.Thisannulusisdrawn lines correspond to the “local” backgrounds extracted from the same in Fig. 2. However,the number countsin this circle are higher imagesasourclusters(red:intheannulusbetween thetwocirclesof thantheCFHLTSDeepsurveycounts(seebelow),probablybe- Fig.2,blue:inthenorthrectangle,seetextandFig.2).Errorbarsare cause the annulus is too close to the clusters and still includes Poissonianandarenotplottedforclarity. clustergalaxies,sowewillnotusethesecountsasabackground. The Canada-France-Hawaii Telescope Legacy Survey theGLFsthusderivedforAbell222and223.Thiswillillustrate (CFHTLS)hasbeentakenwiththesametelescope,cameraand thedifficultytodetermineGLFparametersunambiguously. filters as the data we are analyzing. The Deep survey explores a solid angle of 4 × 1 deg2 of the deep Universe, in four in- We can note from Fig. 5 that galaxy counts are notably dependentfields(http://www.cfht.hawaii.edu/Science/ higherinAbell223thaninAbell222,implyingthattheformer CFHLS/).Observationsarecarriedoutinfivefilters(u∗,g(cid:4),r(cid:4),i(cid:4) clusterisricherthanthelatter. andz(cid:4))providingcatalogsofsourcesthatare80%completeup toiAB =26.0(seehttp://terapix.iap.fr/cplt/oldSite/ 3.3.Galaxyluminosityfunctions Descart/CFHTLS-T0005-Release.pdf). We only consider the Deep survey here, because we are interested in the GLFs The Galaxy Luminosity Functions (GLFs) of Abell 222 and downtofaintmagnitudes.Forthefourdeepfields,wecomputed Abell223werecalculatedinbinsof0.5magandnormalizedto the galaxynumbercountsin binsof0.5mag,ing(cid:4) andr(cid:4), nor- 1deg2,asdescribedabove.Wesubtractedthebackgroundcon- malizedtoa surfaceof1 deg2. Thegalaxycountsinthese four tribution using as backgroundgalaxy counts: 1) the average of fields are drawn in Fig. 5. As discussed before (see e.g. Boué thethreeDeep1,Deep2andDeep3fields;2)the“local”counts etal.2008,andreferencestherein),thecountsinthesefourfields inthenorthbluerectangleofFig.2. somewhat differ, due to cosmic variance, and in particular the The GLFs are displayed for Abell 222 and Abell 223 in countsintheDeep4fieldarehigherthanintheotherthree.We Figs.6and7respectively.Theerrorbarsdrawninthesefigures will thereforetake as backgroundgalaxycountsthe averageof weretakentobe4timesthePoissonianerrorsongalaxycounts, theDeep1,Deep2andDeep3fieldcounts. as suggested by detailed simulations previously performed by ourteamforsimilardata(seeBouéetal.2008,Fig.5). For comparison, we also extracted galaxy counts from a TheGLFs(asafunctionofabsolutemagnitude)werefitby region assumed to be representative of the background in our aSchechterfunction: image. This region was chosen to be a rectangle north of the clusters, in an area devoidof brightstars. It is shownas a blue S(M)=0.4 ln10φ∗yα+1e−y rectangleinFig.2andcoversaneffectiveareaof0.06825deg2. The corresponding counts (hereafter the “local” background withy=100.4(M∗−M). counts) appear to be notably lower than the CFHTLS-Deep TheparametersoftheSchechterfunctionfitsoftheGLFsare counts(Fig. 5); this may be the case if this regioncorresponds respectivelygiveninTables2and3 forthetwo differentback- toavoidlocatedbetweenthelargescalefilamentsthatconverge groundsubtractions:theaverageoftheCFHTLSDeepfieldsand towardstheclusters. the “local” background.The absolute magnituderange consid- AlthoughweconsiderthattheCFHTLS-Deepcountsproba- eredisindicatedforeachfit.ThefitsaredrawninFigs.6and7 blyrepresentbetterthegalaxybackgroundcounts,wewillalso forAbell222andAbell223respectively. fit the GLFs with these “local” backgroundcounts, keeping in If we compare the results given in Tables 2 and 3, we see mindthefactthatthese“local”countsarelikelytooverestimate thattheΦ∗ and M∗ parametersdependlittle onthebackground Page5of15 A&A517,A94(2010) Table2.Schechterparametersforgalaxyluminosityfunctions(backgroundgalaxycountstakenfromtheCFHTLSDeepfieldcounts). Cluster Filter Range Φ∗ M∗ α Abell222 g(cid:4) [−22.5,−16.0] 2891±134 −20.04±0.06 −0.60±0.05 r(cid:4) [−23.5,−16.0] 937±64 −22.01±0.07 −1.43±0.01 Abell223 g(cid:4) [−25.5,−15.5] 1017±72 −21.05±0.06 −1.30±0.02 r(cid:4) [−26.0,−15.0] 1107±62 −22.05±0.06 −1.29±0.01 Table3.Schechterparametersforgalaxyluminosityfunctions(“local”backgroundgalaxycounts). Cluster Filter Range Φ∗ M∗ α Abell222 g(cid:4) [−22.0,−16.0] 2269±133 −20.32±0.06 −0.94±0.03 r(cid:4) [−23.0,−16.0] 1132±63 −21.85±0.06 −1.42±0.01 Abell223 g(cid:4) [−25.0,−15.0] 819±55 −21.20±0.05 −1.42±0.01 r(cid:4) [−26.0,−15.0] 1042±58 −22.10±0.06 −1.34±0.01 Fig.6. Galaxy luminosity functions for Abell 222 in the g(cid:4) (top) and Fig.7.SameasFig.6forAbell223. r(cid:4) (bottom)bands,inlogarithmicscale.Thepoointscorrespondtothe subtraction of the background counts taken from the CFHTLS.Error barsare4timesthePoissonianerrorsongalaxycounts(seetext).The ForAbell222,theGLFisquitewellfitbyaSchechterfunc- bestSchechterfunctionfitsaredrawninredforthesubtractionofback- tionin ther(cid:4) band,witha faintendslopeα = −1.43(thesame ground counts taken from the CFHTLSand in blue for “local” back- forbothbackgroundsubtractions).Ontheotherhand,theGLF groundcounts(seetext). is surprisinglyflat in g(cid:4) (slope α = −0.60or −0.94,depending onbackgroundsubtraction). A Schechter function is obviously not sufficient to fit the galaxy counts chosen. On the other hand, some variations are GLFsofAbell223,bothing(cid:4) andr(cid:4):Abell223hasmorevery found in the faint end slope α, as expected since it is at faint brightgalaxiesthanAbell222,andasecondcomponentatbright magnitudesthatthebackgroundgalaxysubtractionhasastrong magnitudesisobviouslyrequired.Atfaintmagnitudes,theGLF influence. slopesing(cid:4) andr(cid:4) arecomparableforAbell223.Theyvaryby Page6of15 F.Durretetal.:Abell222/223 3.4.Morphologicalsegregation We applied a new tool developed in SExtractor that calculates foreachgalaxytherespectivefluxesinthebulge(spheroid)and xtractor disk. This new experimental SE feature fits to each galaxyatwo-dimensionalmodelcomprisedofadeVaucouleurs spheroid(bulge)andanexponentialdisk.Briefly,thefittingpro- cess is very similar to that of the GalFit package (Peng et al. 2002)andisbasedonamodifiedLevenberg-Marquardtminimi- sation algorithm.The modelis convolvedwith a supersampled modelofthelocalPointSpreadFunction(PSF),anddownsam- pled to the final image resolution. The PSF model used in the x fitswasderivedwiththePSFE software(Bertinetal.2010,in preparation)fromaselectionof5436(g(cid:4))/6922(r(cid:4))pointsource images.ThePSFvariationswerefitusinga6thdegreepolyno- mialofxandyimagecoordinates. Themodelfittingwascarriedoutindependentlyintheg(cid:4)and r(cid:4) bands.Thebulgetototalratiosdiscussedherewereextracted fromtheredchannel,limitedtor(cid:4) =22,sincebeyondthismag- nituderesultsmaybecomeunreliable. We applied this tool to look for evidence for morphologi- cal segregation, by computing the fraction of galaxies with a spheroid to total flux smaller than 0.3, as a function of radius Fig.8.Fractionofdisk-dominated galaxiesasafunctionofradiusfor (i.e.inconcentriccirclesofstep0.02deg).Theresultisplotted Abell222(inmagenta)andAbell223(incyan). inFig.8.Ineachbin,1σerrorswerecomputedasfollows:ifn isthenumberofgalaxieshavingaspheroidtototalfluxsmaller than 0.3 and N the total number of galaxies, we assume that n and (N − n) are independent variables; in this case, the error onn/N is: smallquantities(0.12and0.05respectively)betweenonegalaxy (cid:2) backgroundsubtractionandtheother.However,theseslopesare (n/N) (1/n−1/N). only indicative, since as seen in Fig. 7 the GLFs show strong wigglesforabsolutemagnitudesbetween−20and−17inr(cid:4)and between −20 and −16 in g(cid:4). Since Abell 223 is itself a double Wecanseethatthereisageneralincreaseofthefractionofdisk- cluster (Dietrich et al. 2002), it is not surprising to find that it dominated galaxies with radius as expected (see e.g. Biviano hasa“perturbed”GLF.Thisisalsothecaseforexampleforthe etal.1997). Coma cluster, known to have a two-component GLF (Biviano Inordertoseeiftherewasamorphologicalsegregationthat etal.1995). could be linked to the presence of cosmological filaments, we IfwecomparetheGLFsofthetwoclusters,wecanseethat computedthatsamefractioninanglesof45degallaroundeach theyare moreor lesscomparablein ther(cid:4) band(exceptforthe cluster.TheresultisshowninFig.9,whereangularsectorsare wigglesin Abell223),whilein the g(cid:4) bandthe GLFis notably numbered from 1 to 8 clockwise from east (see Fig. A.1). No steeperin Abell223thanin Abell222.Anexplanationforthis strong trend is foundin this plotwithin error bars, as expected can be that both clusters have comparable old galaxy popula- if there is no obvious cosmological filament linking the clus- tions,butthatthenumberoffaintstarforminggalaxiesishigher ters with the surrounding cosmic web. We can note however in Abell 223, where star formation can be triggered by recent that for both clusters there is a dip in the fraction of bulge- interactions. dominated galaxies in sector 6, which corresponds to the line joiningthetwoclusters;thefilamentalreadydetectedbyWerner If we compare these GLFs to those found by other au- et al. (2008) using wavelet-decomposition in the [0.5−2] keV thors, we can note that the bright part of the GLF Schechter energybandislocatedalongthissamedirection.Inthiszonethe fits (r(cid:4) < 18, or Mr(cid:4) <−18.9) for both clusters are very simi- gasisnotyethotenoughtoinfluencethegalaxies.Ifwelookat larinshapetotheGLFsrecentlyobtainede.g.byAndreonetal. Fig.9fromLaganáetal.(2009),2keVisthelowerlimitwhere (2008);theseauthorsanalyzedtheGLFsofasampleofclusters ram-pressure plays an important role in galaxy clusters. Thus atvariousredshifts,limitedtorelativelybrightabsolutemagni- wewouldexpecttodetectatrendonlyifthefilamentwashotter tudes:M <−19. V than2keV. TheαslopesofthefaintendsoftheGLFsderivedhere(ex- Another possibility for this dip in the fraction of bulge- cept for that of Abell 222 in the g(cid:4) band) are within the broad dominatedgalaxiesisahighernumberofbluegalaxiesinwhich rangeofvaluesestimatedbypreviousauthorsfordifferentclus- starformationwastriggeredduetoongoingcollisions.Poggianti ters,clusterregionsandphotometricbands(seee.g.thecompi- et al. (2004) arguedthat there is a striking correlationbetween lationinTableA.1ofBouéetal.2008). thepositionsoftheyoungandstrongpost-starburstgalaxiesand NotethatAbell222andAbell223are atredshift0.21,and substructure in the hot intra-cluster medium (ICM) identified few GLFs are available for clusters at such redshifts. Andreon from XMM-Newton data, with these galaxieslying close to the etal.(2005)foundsomewhatshallowerslopesforthreeclusters edgesoftwoinfallingsubstructures.Thisresultsuggeststhatthe atredshifts∼0.3,butintheK band,sothecomparisonwithour interactionwiththedenseICMcouldberesponsibleforthetrig- resultsisnotstraightforward. geringofthestarformation. Page7of15 A&A517,A94(2010) Fig.11. Positionsof the galaxieswith measured redshifts and magni- tudesintheregionscoveredbyAbell222andAbell223(blackcrosses). Thegalaxiesbelongingtothetwomaingravitationallyboundsystems Fig.9.Fractionofdisk-dominatedgalaxiesinsectorsof45degwithin identifiedwiththeSGmethodareshownincolours(seetext). each cluster,withthesamecolours asintheprevious figure.Angular sectorsarenumberedfrom1to8clockwisefromeast(seeFig.A.1). 4. Cluster-scaletolarge-scalestructure WesearchedtheNEDdatabaseforgalaxieswithredshiftsavail- ableinaverylargeregionof3degradiusaroundAbell222/223 and found 980 galaxies. This radius corresponds to 37 Mpc at thedistanceofthestudiedclusters,i.e.tothetypicaldiameterof a“bubble”orofavoidintheuniverse(seee.g.Hoyle&Vogeley 2004).Thiscatalogueofgalaxieswithmeasuredredshiftswillbe usedfirsttosearchforsubstructuresinthezonescoveredbyour clusters,andsecondtosearchforstructures(suchasfilaments) atalargerscale. 4.1.Searchforgravitationallyboundstructures Within the regions of our images covered by Abell 222 and Abell223,246galaxieshavemeasuredredshifts,and210galax- ies haveboth measuredredshiftsandmagnitudes.Outof these 210galaxies,173haveredshiftsinthe[0.18,0.24]interval.The positionsofthe210galaxieswithmeasuredredshiftsandmag- nitudes are displayed in Fig. 11. It is interesting to note that, as traced by these galaxies, both clusters appear to be strongly elongated, with elongationposition angles that differ from one cluster to the other. This strongly suggests that the clusters are Fig.10.Averager(cid:4)magnitudeasafunctionofradiusforAbell222(in notindynamicalequilibrium,inagreementwithpreviousstud- magenta)andAbell223(incyan). ies (see above, and also Zabludoff et al. 1995; Roettiger et al. 1996;Boschinetal.2004;Girardietal.2006). Thecorrespondingredshifthistogram,zoomedonthecluster redshiftinterval,isshowninFig.12. 3.5.Luminositysegregation We applied to the catalogue of 210 galaxies with redshifts andmagnitudestheSerna&Gerbal(1996)algorithm(hereafter We searched for luminosity segregationin both clusters by se- SG), which allows to search for substructures, separate galax- lecting galaxies brighter than r(cid:4) = 24 and within the red se- ies forming gravitationally bound structures, and estimate the quence defined in Fig. 4, to increase the probability that they masses of these substructures. The dendogram thus obtained belongtotheclusters.Themeanr(cid:4)magnitudewascalculatedin is shown in Fig. 13. It clearly shows two dynamically dis- eachoftheradialbinspreviouslydefinedandresultsareshown tinctsubstructures,correspondingtothetwoclusters:Abell222 inFig.10.Weobserveanincreaseofthemeanmagnitudewith in the top half and Abell 223 in the bottom half. The two radius: it is about 21.3 in the central bin, and then increases maingravitationallyboundstructurescorrespondingtothesetwo withradiusfromvaluesaround22to22.6. clusters respectively include 55 and 64 group members. Their Page8of15 F.Durretetal.:Abell222/223 Fig.12.Redshifthistogramsforallthegalaxieswithmeasuredredshifts intheregionscoveredbyAbell222andAbell223(inblack),andfor the galaxies in the gravitationally bound structures of Abell 222 and Abell 223, in magenta and cyan respectively (the positions of these galaxiesareshowninFig.11)withthesamecolourcoding. corresponding velocity dispersions are 1014 and 1170 km s−1 respectively. The virial masses derived for these two substruc- tures (assuming M/L = 400 solar units) are 1.9 × 1011 and 4.2×1013 M(cid:5). ForAbell222,andprobablyalsoforAbell223,thesemasses are obviouslyunderestimated,particularlyfor Abell 222, since Fig.13.DendogramshowinghowAbell222andAbell223canbedy- such a valuedoes notevenreach that of a cD galaxy.We have namicallyseparatedwhenapplyingtheSerna&Gerbalalgorithm.The triedtoanalyzethetwoclustersseparately,butthismodifiesthe abscissaisthenegativebindingenergywhilethecataloguenumbersof resulting masses only slightly. The fact that we find too small thevariousgalaxiesareshownalongtheordinate(thetopofthedendo- cluster masses can be due to severalreasons. First, the redshift gramistruncated,butwouldonlyshowaconstantgravitationalbinding sample we are using is taken from the NED data base, and is energy). obviouslynotcomplete,evenatbrightmagnitudes.Wehavees- timatedthecompletenessoftheredshiftcataloguebycomparing thenumbersofgalaxieswithredshiftsinthe[0.18,0.24]redshift simulations that for merging systems where the subcluster has intervaltothetotalnumberofgalaxies,inmagnitudebinsof0.5 amasslargerthanonefourththatofthelargecluster,thevirial inther(cid:4) band.Wefindthatfor17<r(cid:4) <19.5thecompleteness masscanbelargerthantwicetherealmass.Thismostprobably is around 50%, then rapidly decreases for r(cid:4) > 19.5, confirm- appliesatleasttotheAbell223cluster. ing that the cluster masses based on such a redshift catalogue We can also apply the mass-temperaturerelation estimated are probably unreliable. Second, as pointed out by the referee, from X-raydata by Arnaudet al. (2005), with the overalltem- the SG method relies on the unlikely assumption that galaxies peratures given in Table 5; this relation gives masses M of 200 inclustersarethemasscarriers,andadditionallyrequiresanas- 3.8×1014 and 4.7×1014 M(cid:5) for Abell 222 and Abell 223 re- sumptionabouttheM/Lratio,heretakentobe M/L=400;this spectively(alsowithH =70kms−1Mpc−1).Thesevaluesmay 0 valuemaybereasonableforaclustertakenasawhole,butistoo beclosertotherealvaluesthanthoseestimatedfromthegalaxy largeforindividualgalaxies(thedarkmatterhalosofindividual velocitydispersions,sincee.g.Takizawaetal.(2010)foundthat galaxies would be overlapping in the dense central regions of for merging clusters X-ray derived masses were usually more clusters,andstrongmasssegregationwouldresultbydynamical reliablethanvirialmassestimations.However,asexplainedfor friction,whichisnotobserved). example by Nagai et al. (2007), although the total ICM mass Better estimates of the cluster masses can be obtained by can be measuredquite accurately (to better than ∼6%) in clus- applying other methods. For example, the relation between ters, the hydrostaticestimate of the gravitationallyboundmass massandvelocitydispersioncomputedbyBivianoetal.(2006) isbiasedlowbyabout5−20%throughoutthevirialregion,pri- leads to masses of 1.2 × 1015 and 1.4 × 1015 M(cid:5) (with marilydueto additionalpressuresupportprovidedbysubsonic H = 70 km s−1 Mpc−1) for Abell 222 and Abell 223 re- bulkmotionsintheICM,ubiquitousinoursimulationsevenin 0 spectively,basedonrespectivevelocitydispersionsof1014and relaxedsystems.Anothersourceoferroristhefactthattheclus- 1070km s−1. Note howeverthat since the clustersare notviri- tertemperatures,particularlythatofAbell223(seeFig.18)are alized these values are probably overestimates. For example, nothomogeneous.Forexample,Rasiaetal.(2004)haveshown Takizawa et al. (2010) have recently shown from numerical thatanincompletethermalisationofthegasduetoanincomplete Page9of15 A&A517,A94(2010) Table4.InformationonX-raydata. Name Filter t (ks) n (1020cm−2) exp H MOS1 MOS2 pn Abell222 Thin1 11.814 18.005 10.226 2.26 Abell223 Thin1 23.479 26.069 10.612 2.26 excluded substructure PN -12:48:00.0 -13:00:00.0 12.5 arcmin Fig.14. Distributionofthegalaxieswithavailableredshiftsinalarge regionsurroundingtheclusters.Bluepoints:allgalaxieswithmeasured rTseohduesthhtwiwftoess,ctr:leuAdstbpeerolslin2ut0ns9:d.egrTalshatexuipdehysyawsrieictahclirrsecidzleesdho.iffAttshtiihnsimrtdhaecp[liu0ss.14te84r,.70ap×.2p44e4]a.ri7nstMteorpvtcah2le.. 12:00.0 9:12.0 48.0 24.0 1:38:00.0 37:36.0 37:12.0 48.0 24.0 36:00 Fig.15.Annuluswherethebackgroundwasextractedforsubtraction. virialisationoftheclustercouldleadtoanunderestimateofthe mass.Ontheotherhand,iftheX-raygasisheatedbyacollision, themasscouldbeoverestimated. Table5.OverallX-raytemperaturesandmetallicities. WecanalsonotethatMamon(1993)hascalculatedthebias onthe M/Lratiowhenthevirialtheoremiswronglyappliedto Cluster r kT Z 200 agroupwhichisstillevolvingdynamically(seehisFig.11).He (Mpc) (keV) (solar) found that in some cases the mass could be underestimatedby Abell222 1.28±0.11 3.77±0.15 0.23±0.06 a factor reaching 100. Therefore the masses estimated for our Abell223 1.55±0.15 4.38±0.16 0.23±0.08 clustersassumingthattheyarevirializedareclearlynotvalid. Unfortunately, the SG method does not allow to separate thetwosubstructuresinAbell223discoveredbyDietrichetal. v8.0andcalibrationdatabasewithallupdatesavailablepriorto (2002)basedongalaxydensitycontours(seetheirFig.6). November 2009. The initial data screening was applied using Note that Abell223 is probablyclose to the intersection of recommendedsetsofeventpatterns,0−12and0−4fortheMOS twoormorefilaments,sinceitismassiveandsubstructured,and andPNcameras,respectively. itsGLFiscomparabletothate.g.ofComa(seeSect.3.3).Onthe Thelightcurvesare notconstantandlargevariationsin in- otherhand,Abell222isprobablyonlyfollowingitspathalong tensity are visible (flares). To improve the signal-to-noise ratio afilament. wediscardedperiodsofflaresandthecleanedlightcurvesinthe energyrangeof [1−10]keV exhibitedstable meancountrates; 4.2.Large-scalestructure exposuretimesaregiveninTable4.Weconsideredeventsinside thefieldofview(FOV)andexcludedallbadpixels. Out of the 980 galaxies found in NED in a 3 deg radius re- ThebackgroundwastakenintoaccountbyextractingMOS1, gion, 310 have redshifts in the [0.18,0.24] interval (i.e. about ±9000 km s−1 around the mean cluster velocities). Their spa- MOS2 and pn spectra from the publicly available EPIC blank skytemplatesofAndyRead(Read&Ponman2003).Theback- tialdistributionisshowninFig.14.Aclusterisvisibletowards ground was normalized using a spectrum obtained in an annu- the southwest: Abell 209. It has a redshift similar to that of lus(between12.5−14arcmin)wherethecluster emissionisno Abell 222 and Abell 223 and is located about 19.2 Mpc away longer detected. In one of the pointings, the A223 cluster falls fromAbell222inprojectiononthesky.Nosignificantfilamen- tarystructureisdetectedtowardstheAbell222/223clusters. exactlyinside theannuluswherewe determinethe background contribution. We thus masked this region, avoiding any cluster contamination. In Fig. 16 we show a comparison between the 5. X-rayanalysis observedandReadbackgroundsforthethreedetectors,aswell astheresidualsbetweenbothdeterminations. 5.1.Datareduction Inordertodeterminetheglobalpropertiesoftheseclusters, The pair of clusters Abell 222/223 was observed with weadoptedthevirialradiir determinedbytheweaklensing 200 XMM-Newton in two different pointings(revolutions1378 and analysis (Dietrich et al. 2005). These values, together with the 1380)with a totalexposuretime of144ks. We useddatafrom mean temperaturesandmetallicitiesare givenin Table5. They all EPIC cameras (MOS1, MOS2 and pn). The data were re- werecalculatedfixingthehydrogencolumndensitytoitsgalac- duced with the XMM-Newton Science Analysis System (SAS) ticvalueofn =1.56×1020cm−2. H Page10of15